Input variable selection for model-based production control and optimisation

ProOpter: An advanced platform for production analysis and optimization

“…[28], and further elaborated and extended in Glavan et al [10] and Glavan et al [9]. The concept is schematically shown in Fig.…”

“…Different variable-ranking methods are contained in the IVS module, such as: (partial) Correlation, (partial) Mutual Information, Gamma Test, ANOVA, NonNegative Garrote, LASSO, etc. For a detailed discussion and an evaluation of the integrated methods, see Glavan et al [9]. Note, that the list of supported methods is not strictly closed, as new methods can easily be added to the module.…”

“…We used the system that was stabilized with the low-level control presented in [19], where nine outputs are controlled with cascade loops. In order to realize the HPC concept some changes in low-level control were added, as suggested in [9]. Therefore, ten input variables are available to manipulate the TE process.…”

“…We can build a model using neural-network modelling techniques [9], or using a fuzzy modelling approach [6]. In Fig.…”

“…With this in mind we have introduced and further elaborated the concept of holistic production control (HPC) (see [9,10,28]), which relies on the well-known, modelbased, process-control paradigm. However, the developed ideas and methods are of little value if they are not supported by an appropriate tool.…”